International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 07 Issue: 04 | Apr 2020
p-ISSN: 2395-0072
www.irjet.net
Review of Pest Attack Prediction and Detection Methodologies Swarag Gutte1, Juhilee Nazare2, Sneha Pawar3, Shweta Karanjawane4, Prof. Shobha Raskar5 1-5Department
of Computer Engineering, Modern Education Society’s College of Engineering, Maharashtra, India ---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - The pests and diseases in the fields are the major contributors to the crop loss. In countries like India with a large population, it is necessary to get maximum outcome out of farming and that can be possible if we can control the pest attack in the fields. In this digital world, many advanced technologies are available to fight the pest attack. By using these technologies, we can predict the pest attack, determine which pest has attacked and can take the required measures in advance to reduce the loss. These technologies include machine learning, artificial intelligence, computer vision, deep learning and many more. In this paper, we will be looking at some promising methods to be implemented by the researchers to counter the pest attack in the fields.
Due to this, farmers cannot use one kind of pesticide for all the crops, and if they use the wrong pesticide for some crops it won’t give effective results and will lead to economical loss and crop yield. Early detection of pest attacks or proper detection of the pest attack is necessary. In this paper, we will be having a look at some of the pest detection methods. These methods use computer vision, machine learning or deep learning. Many methods use these technologies individually for giving the results but in recent times integrations of two or more such technologies are under study which will give more accurate results. 1.1 Automatic counting the Insects/pests in the field using the computer vision
Key Words: Pest attack, Machine learning, Artificial network, Neural network, Computer vision, Expert system.
In this approach, the number of pests on any plant are automatically counted without any human interference in the process. The insect population growth control is based on the infestation level and plant’s development stage. Most of the times, this is done manually which is time consuming and hence there is growing motivation for using digital images collected directly from the field, making it possible to develop a computer vision system to identify and count different species of pest and insects from the field. For implementing the system, it is necessary to collect a large sample of dataset in the form of digital images automatically, hence unmanned aerial vehicles can be used which will have the cameras mounted on them to capture the images continuously.
INTRODUCTION Since ages agriculture is an essential part of our lifestyle. But in the current scenario because of rapid increase in population the agricultural land area has been reducing as that land is converted into residential areas, but on the other hand the population is increasing at a similar pace. In addition to this climate change is also affecting the yield in the fields. Hence smart agricultural methods are required in the countries like India where there is shortage of agricultural land area. When compared with population so “more with less” concept is followed. In any farm the biggest problem for farmers are the pests which damage the yields completely and which leads to the loss of crops and economic loss. To overcome this problem farmers, use heavy pesticides to kill the pests in the fields, these pesticides are strong chemicals and are harmful for crops as well as the people consuming it in the form of the agricultural product.
In stage 1, image segmentation is done by simple linear iterative clustering superpixels method to segment the images collected as per the pest’s categories. For example, the set of 10000 collected images can be divided into groups of 7 clusters to increase the accuracy and training for the deep learning model in the classification. It is trained with the three different types of approaches to increase the accuracy, they are fine tuning, Image net, transfer tuning.
Most of the farmers use pesticides just to be safe and not on the basis of any strong evidence of the attack or just because someone in the neighboring land has used the pesticide. This leads to unnecessary use of pesticides hence affecting the crops as well as the lives of farmers. Therefore, only required amount of pesticides should be used by farmers to reduce the harmful causes of pesticides on humans and crops, but at the same time they cannot risk the damage to the plants by these pests. Agricultural pest is defined as the insects/animals feed on the plant tissues.
In deep learning, designing the automatic counter to count the eggs (soyabean cyst) on the plants using the microscopic images. These images are trained by the convolutional neural network with data labelled to learn how to reconstruct an egg pattern converting the image size to patches of 16×16 pixels to determine if the patch contains an egg or not. The identified eggs are then counted using the matrix labelling function. Similar method is applied in the pakchoi leaves. In this method, different stages of aphids are obtained, In this way, using binary mask over each image
These pests are of different types and different plants are affected by different types of pest attacks. These pests mainly depend on the weather conditions and crops.
© 2020, IRJET
|
Impact Factor value: 7.529
|
ISO 9001:2008 Certified Journal
|
Page 5948